The hardest part of building OTP is not building OTP. It is the hour I spent yesterday explaining transactive memory to a guy on LinkedIn who was nodding at every wrong word.
He kept saying "so it's like a prompt library." I would say no, it is queryable cross-org memory. He would say "right, like a shared prompt library." I would say no, prompt libraries cap at their token budget, this is selective retrieval that grows with network participation. He would say "right, like a really big prompt library."
I closed the tab. I went and got coffee. I came back and he had sent a paragraph that started with "so to summarize, your tool is."
This is a specific kind of struggle that does not get written about because it sounds like sour grapes when you write it. So let me write it.
When you build something five years too early, the first hard part is not finishing it. It is finding the people who can see it. Not customers. Not even early adopters. People whose mental models already have the shape of the thing you built, so when you describe it they recognize it instead of pattern-matching it onto something they already understand.
I have shipped OTP for five weeks. The number of people who have seen what it is, on first description, is four. Two of them I had already met. One of them is a founder named Dick Chie who I did not solicit. He sent me an unprompted message this week using the academic phrase for the thing I had been trying to describe in marketing language. He had no stake. He saw it. The fourth is Sean Donahoe, who is building two adjacent protocols at the same layer of the stack and recognized OTP as the third leg.
Four people in five weeks. That is the population of the bleeding edge for this idea, today.
The wrong response to that is to argue harder with the LinkedIn guy. The right response is to find the next four. Then the next four. Then keep cutting friction at the door for the people who arrive after them.
I am not tired. I am calibrated. The work is the calibration.
I am back at the cold outbound this morning. The list is forty names. The job is to find the one who reads "queryable transactive memory across organizations" and says "oh, that is the thing I have been waiting for."
The other thirty-nine will say it is a prompt library.
That is fine. They are not the four.